Power Query gave me the wrong number


Power Query lies to you. Not always. Sometimes. And it means well.

It wants to get you results faster. But sometimes speed gets in the way of accuracy.

I ran a query this week to get the latest stock count per product. Power Query gave me 49. The actual number was 8.

No error message. No warning. Just the wrong answer.

And this isn't just a stock count thing. It can happen any time you're trying to keep the latest record per group - latest order status, latest price update, latest anything.

If you use Power Query for this, you should see this.

video preview

Power Query Challenge Pack - available now

This is actually one of the challenges in our just-released Power Query Challenge Pack. (There's a whole other way to solve it, so you can still challenge yourself.)

15 real-world problems. Messy data. No hand-holding. You work through it, get stuck, figure it out. Only then check your approach against a video solution.

Every challenge came from real feedback. What professionals actually find hard at work. Duplicate handling, broken queries, PDFs, files that were built for presentation and not for analysis.

🤓 Geeky News

🤖 From Copilot to Autopilot: the AI Agent that runs itself

Microsoft just introduced something called Autopilots. Always-on agents that work in the background, take action on your behalf, and don't wait to be asked.

Their first one is called Scout. It connects to Teams, Outlook, OneDrive, SharePoint, your calendar, your chats. It schedules meetings, blocks focus time, flags stalled decisions, preps your briefing materials. All without you prompting it.

The pitch is that it learns how you work over time and gets more useful as it goes.

Currently in private preview for a select group, with Frontier program access for early adopters.

Whether you want an AI that knows your schedule, your priorities, and your contacts running quietly in the background all day is, of course, a personal decision. 👀

📧 Latest improvements in the New Outlook

Microsoft posted a roundup of 15 productivity features that have landed in the new Outlook for Windows over the past year. Most of them are small. A few are genuinely useful.

The ones worth knowing:

  • you can now snooze emails to bring them back when they're relevant,
  • pin important messages so they stay at the top of your inbox,
  • Sweep lets you set automatic cleanup rules for specific senders,
  • follow a meeting instead of declining it to get the recap without having to attend,
  • as well as more customization settings.

If you've been putting off switching to the new Outlook, the gap is narrowing. I'll hold off for a while longer.

💻 Office 2019 for Mac stops working July 13th

If you're still on Office 2019 for Mac, the apps will switch to read-only mode on July 13th. You'll be able to open and print files, but not edit them or create new ones. Office 2019 for Windows is not affected.

Options if you are affected: upgrade to a Microsoft 365 subscription, buy Office 2024 as a one-time purchase, or use the free Office web apps at office.com.

🤔 Did You Know?

A new Copilot in Excel feature is looking promising.

You can now personalize the default behavior and save those settings across workbooks. It's like custom instructions in ChatGPT, but for Excel.

List your preferred functions and formatting styles.

Copilot will remember, so you don't have to mention them in every prompt.

👏 From certified to actually using it

Nikoleta works in financial analysis. The kind of role where data doesn't arrive clean, reporting deadlines don't move, and "I'll figure it out later" isn't really an option.

She just finished Master Excel Power Query: Beginner to Pro and posted this on LinkedIn:

Congrats, Nikoleta. Certificate earned, and already being used at work. That's the order you want it in.

See you next week,

Leila

When you're ready, here are some ways we can help:

🎓 Join 400,000+ members in our courses

📺 Get free tutorials on YouTube

👥 Train your whole team with our courses. Team pricing and progress tracking - reply for details.


This newsletter contains affiliate links, which give us a small commission at no cost to you. Thank you for your support!

Leila Gharani - XelPlus

XelPlus is a leading online education company, providing training courses for Excel, Power BI, Finance, and Google Sheets. XelPlus’ bestselling courses are popular among financial analysts, CFO’s, and business owners. Technology is changing fast. We help our members turn confusion into confidence with every skill learnt.

Read more from Leila Gharani - XelPlus

I once had a task that sounded really simple. I had a list of names and a list of months like the below: I needed every name paired with every month. Like this: Easy enough, right? Except doing it manually is one of those things that sounds quick and then just... isn't. Especially if you have 12 months and a whole bunch of names. And what if someone adds a name in the middle? Yeah, you don't want to do it manually. It will be an ongoing headache. Power Query does it in a few clicks. No...

At my old job, messy dates were a constant headache. People would overwrite my data validation with whatever format they felt like. SAP exports came in with dates Excel refused to recognize. I tried reformatting the cells. Nope. That didn't help. So I added a check to catch the problem cells. Unfortunately, there were hundreds of them. And then I got to work, fixing each by hand. Today I'd fix the whole column in 10 seconds. One Python formula. Written in a cell, just like any Excel formula....

A German Controller Magazine recently said certain time series analysis techniques are "not realistically feasible with Excel." I read that and thought: challenge accepted. So that's what I opened with at the Global Excel Summit in London on Tuesday this past week. Three techniques. Python's data science libraries. Done right inside Excel, live on stage. (And no, the cape on the table isn’t mine. IYKYK) Because Python is in Excel now. The limits people keep assuming are there... a lot of them...